Taste is still the job: give creative direction a reviewable standard

Generating more variations does not decide which work belongs in a game. Someone still has to choose a direction, reject attractive mismatches, and explain how the accepted work serves the Player Promise.

How We Ship gives that judgment a reference: the approved Vertical Slice becomes the Quality Baseline. The complete game must meet or improve on its coherence and polish. This position applies that rule to creative direction; it does not claim measured creative results.

Write a direction that can guide a decision

Connect the work to a player-facing purpose. Name the intended tone, what must remain readable, and what the work needs to share with neighboring assets or scenes. Keep examples available so the next contributor can evaluate the same constraints.

For a hypothetical warning icon, readability at gameplay size can matter more than detail in a large export. For a dialogue scene, a fluent paragraph can still conflict with the character's established intent. These are examples of review questions, not Clowdr project findings.

Generation, stock material, and hand-authored work all need this direction. The source method does not establish fit.

Check the work where players encounter it

Review visual work at its intended size, against the game's backgrounds and motion. Check whether important states remain distinguishable and whether adjacent assets follow a coherent direction. Read writing in its scene, including the choices and consequences around it.

When the direction changes after the slice, record why. Show how the revision still meets the Player Promise and how the complete work compares with the baseline. Quietly accepting lower coherence because many assets already exist undermines Craft Proof.

A reviewer should be able to point to a specific mismatch and request a specific revision. “I dislike AI” and “it looks polished” both leave the production decision unexplained.

Keep creative responsibility with the contributor

The human owner must be able to explain the choices, revise the result, account for its source, and support the handoff. Record material sources and rights concerns, including unresolved imitation or attribution questions, before accepting proposed final work. This record supports review; it cannot make an unclear source safe by itself.

If the pitch deliberately excludes AI from this area, that is a valid production choice. If it permits assistance, generated variations remain draft material until an accountable contributor can own and verify the accepted work.

Use the Clowdr AI standard for the shared acceptance requirements and the Vertical Slice guide for the baseline. The useful creative record explains what was chosen, what was rejected, and what changed in the game.